Introducing arxiv-sanity

[jaimevarela]

Also, arxiv provides RSS feeds which can be easily parsed. It doesn't take long to write a script which will parse the rss feed and e-mail you papers based on key words or author names. I have such a set up and it's nice to just get updates with what I want to see.
I wish Arxiv would also provide download data on their papers as it would be nice to know what is popular research in various fields.

[MrKeotan]

I recall that Arxiv is not friendly to robots and threatens to ban for automate mass downloading of papers. How do you get over it?

[herp_derpingson]

This is like a youtube video suggestion list for research papers. :D

[habibmrad]

very useful..highly recommended to all ML and DL practitioners...I appreciate your Work @Andrej Karpathy

[AnandMuglikar]

+Andrej Karpathy Isn't it possible to follow interesting users (like u) & pprs they r reading now?

[thntk]

@Andrej Karpathy
This work is great.

I'm quite curious about the performance of SVM for recommendation here. Because there would be a problem of wrong label in training data when a user's library consists only a part of her interested papers. Did you observe the problem?
The de facto solution is computing cosine similarity of paper TF-IDF vector. Is there a reason that you preferred SVM over cosine similarity?

[badme]

i can't open given link of arxiv sanity web sit is it working??

[JadeNagaraja]

This is soo cool! Thanks so much for your work (y)

[daesoolee]

Omg recommendation feature is incredible!

[ilyar]

Such a useful website, but runs out of resources and mostly unavailable lately! :(

This is very interesting! Thanks for this contribution. I'll try to adapt it to my own interests.
I have to make only one observation on the page style: The active tab (I guess "most recent" is the default one) looks *below* the other tabs. I think this can be enhaced by simply reserving the dark bg color for the inactive ones, and white bg for the active one.

[RobertoCamachoB]

A nice feature might be to add folders or some sort of tag for the library, so that if you have papers from different topics you can filter them according to what you are currently interested in.

[jumabekalikhanov]

Cooool, you are my Bro. Right on the spot. I am not only learning science but also how to be good human from you. Wish you all the best for your future.

[marctorsoc]

isn't Google scholar supposed to do this? Anyways, it will be very useful to me as my Google scholar only suggests papers related to my PhD, not related to ML

[DrJimFan]

Something I've been dying to have. Wonderful tool, Andrej!!!

[mrembeh]

If I search for something and I don't have results it just returns white results. You should add a „nothing found“ text please :D

[remusomega]

I can't find any biology papers here, is this only for mathematics based research?

[florijanstamenkovic]

Very interesting work! I am also interested in your choice of training per-user SVMs for suggestions. Training an SVM periodically for each user in a bigram space of features on 17k articles seems a bit expensive... How come you chose that approach over something cheaper (NaiveBayes, vector similarity over extracted features...)?

[abdulqayyumhafiz]

Is there anything to save papers for openreview?

[barath_]

How do I get endorsement to submit a paper. Please help